UNCERTAINTY IN WIND RESOURCE ASSESSMENT - Wind Resource Assessment: A Practical Guide to Developing a Wind Project - page 223

range of uncertainty in practice can be very wide—from as low as 2% in simple,
open terrain to 10% or more in complex terrain—and depends on the model used,
the model's resolution, the terrain and wind climate, the placement of the masts, and
other factors.
Ideally, the uncertainty is determined directly from the on-site wind measurements.
This is possible when the following conditions are met:

• There are at least 5 and preferably 10 or more masts in the project area.
• The masts are well distributed within the proposed turbine array and among the
wind conditions likely to be experienced by the wind turbines.
• There is sufficient data from each mast to accurately compare mean annual wind
speeds.

Here is the procedure. First, one of the masts is designated the reference mast.
Next, the mean wind speed at each of the other masts is predicted from the reference
mast using the wind flow model. The error between the predicted and observed mean
wind speeds is then calculated. The process is repeated with each of the other masts
serving as the reference. Finally, the standard deviation of all the errors is calculated
to estimate the wind flow modeling uncertainty.
More often than not, the conditions listed above are not met. In that case, the
resource analyst must rely on experience (preferably with the same model) at sim-
ilar sites. Proceedings of technical meetings and conferences are a useful reference
for this purpose. An assessment of the complexity of the terrain, variations in land
cover, placement of the masts, and the possible role of coastal sea breezes and other
atmospheric circulations all come into play here.

15.6 COMBINING UNCERTAINTIES

Once the various components of the uncertainty have been defined, they must be
combined in some way to arrive at the uncertainty in the average wind speed of
the entire turbine array. Before this can be done, it is necessary to consider whether
the various sources of error are correlated or uncorrelated . Uncorrelated errors are
independent of one another—the sign and magnitude of one error has no bearing on
the sign and magnitude of another. Imagine a fleet of small boats floating on a choppy
sea. Each boat rides its own wave, its motion unrelated to the motions of the other
boats. Correlated errors, on the other hand, march in lockstep with each other. The
corresponding analogy is a tide that lifts or lowers all the boats at once.
It is better for uncertainties to be uncorrelated because then they do not add in a
linear fashion but rather as the sum of the squares. This reduces the combined uncer-
tainty. For example, suppose the measurement uncertainty and the shear uncertainty
are both 2%, and uncorrelated. The combined uncertainty is
0

. 02 2 + 0 . 02 2 = 2 . 8%